Complexity Certification of the Fast Alternating Minimization Algorithm for Linear Model Predictive Control

In this paper, the fast alternating minimization algorithm (FAMA) is proposed to solve model predictive control (MPC) problems with polytopic and second-order cone constraints. We extend previous theoretical results of FAMA to a more general case, where convex constraints are allowed to be imposed on the strongly convex objective and all convergence properties of FAMA are still preserved. Two splitting strategies for MPC problems are presented. Both of them satisfy the assumptions of FAMA and result in efficient implementations by reducing each iteration of FAMA to simple operations. We derive computational complexity certificates for both splitting strategies, by providing bounds on the number of iterations for both primal and dual variables, which are of particular relevance in the context of real-time MPC to bound the required online computation time. For MPC problems with polyhedral and ellipsoidal constraints, an off-line preconditioning method is presented to further improve the convergence speed of FAMA by reducing the complexity bound and enlarging the step-size of the algorithm. Finally, we demonstrate the performance of FAMA compared to other splitting methods using a quadrotor example.